|
| 1 | +//go:build linux |
| 2 | + |
| 3 | +package main |
| 4 | + |
| 5 | +import ( |
| 6 | + "context" |
| 7 | + "encoding/json" |
| 8 | + "fmt" |
| 9 | + |
| 10 | + "github.com/facebookincubator/go-belt/tool/logger" |
| 11 | + "github.com/mark3labs/mcp-go/mcp" |
| 12 | + "github.com/mark3labs/mcp-go/server" |
| 13 | +) |
| 14 | + |
| 15 | +// ClickUIResult describes the click_ui_element response. |
| 16 | +type ClickUIResult struct { |
| 17 | + MatchedText string `json:"matched_text,omitempty"` |
| 18 | + MatchedResourceID string `json:"matched_resource_id,omitempty"` |
| 19 | + MatchedContentDesc string `json:"matched_content_desc,omitempty"` |
| 20 | + TappedX int `json:"tapped_x"` |
| 21 | + TappedY int `json:"tapped_y"` |
| 22 | +} |
| 23 | + |
| 24 | +func registerClickUIElement(s *server.MCPServer) { |
| 25 | + tool := mcp.NewTool("click_ui_element", |
| 26 | + mcp.WithDescription( |
| 27 | + "Find a UI element by text, resource-id, or content-desc, "+ |
| 28 | + "compute the center of its bounds, and tap it. "+ |
| 29 | + "This is a convenience tool that combines find_ui_element + tap. "+ |
| 30 | + "Returns an error if no matching element is found or if multiple "+ |
| 31 | + "elements match (use more specific criteria).", |
| 32 | + ), |
| 33 | + mcp.WithString("text", |
| 34 | + mcp.Description("Match element whose text contains this substring (case-insensitive)"), |
| 35 | + ), |
| 36 | + mcp.WithString("resource_id", |
| 37 | + mcp.Description("Match element whose resource-id contains this substring"), |
| 38 | + ), |
| 39 | + mcp.WithString("content_desc", |
| 40 | + mcp.Description("Match element whose content-desc contains this substring (case-insensitive)"), |
| 41 | + ), |
| 42 | + mcp.WithDestructiveHintAnnotation(true), |
| 43 | + mcp.WithIdempotentHintAnnotation(false), |
| 44 | + ) |
| 45 | + |
| 46 | + s.AddTool(tool, handleClickUIElement) |
| 47 | +} |
| 48 | + |
| 49 | +func handleClickUIElement( |
| 50 | + ctx context.Context, |
| 51 | + request mcp.CallToolRequest, |
| 52 | +) (*mcp.CallToolResult, error) { |
| 53 | + logger.Tracef(ctx, "handleClickUIElement") |
| 54 | + defer func() { logger.Tracef(ctx, "/handleClickUIElement") }() |
| 55 | + |
| 56 | + textFilter := request.GetString("text", "") |
| 57 | + resourceIDFilter := request.GetString("resource_id", "") |
| 58 | + contentDescFilter := request.GetString("content_desc", "") |
| 59 | + |
| 60 | + if textFilter == "" && resourceIDFilter == "" && contentDescFilter == "" { |
| 61 | + return mcp.NewToolResultError("at least one search filter is required (text, resource_id, or content_desc)"), nil |
| 62 | + } |
| 63 | + |
| 64 | + xmlData, err := dumpUIXML() |
| 65 | + if err != nil { |
| 66 | + return mcp.NewToolResultError(fmt.Sprintf("dumping UI: %v", err)), nil |
| 67 | + } |
| 68 | + |
| 69 | + elements := parseAndFilterUI(xmlData, textFilter, resourceIDFilter, contentDescFilter, "") |
| 70 | + |
| 71 | + switch len(elements) { |
| 72 | + case 0: |
| 73 | + return mcp.NewToolResultError("no matching UI element found"), nil |
| 74 | + case 1: |
| 75 | + // Exactly one match — tap it. |
| 76 | + default: |
| 77 | + data, _ := json.Marshal(elements) |
| 78 | + return mcp.NewToolResultError(fmt.Sprintf( |
| 79 | + "found %d matching elements (use more specific criteria): %s", |
| 80 | + len(elements), string(data), |
| 81 | + )), nil |
| 82 | + } |
| 83 | + |
| 84 | + elem := elements[0] |
| 85 | + cmd := fmt.Sprintf("input tap %d %d", elem.CenterX, elem.CenterY) |
| 86 | + if _, err := shellExec(cmd); err != nil { |
| 87 | + return mcp.NewToolResultError(fmt.Sprintf("input tap: %v", err)), nil |
| 88 | + } |
| 89 | + |
| 90 | + result := ClickUIResult{ |
| 91 | + MatchedText: elem.Text, |
| 92 | + MatchedResourceID: elem.ResourceID, |
| 93 | + MatchedContentDesc: elem.ContentDesc, |
| 94 | + TappedX: elem.CenterX, |
| 95 | + TappedY: elem.CenterY, |
| 96 | + } |
| 97 | + |
| 98 | + data, err := json.Marshal(result) |
| 99 | + if err != nil { |
| 100 | + return nil, fmt.Errorf("marshaling click result: %w", err) |
| 101 | + } |
| 102 | + |
| 103 | + return mcp.NewToolResultText(string(data)), nil |
| 104 | +} |
0 commit comments